Text2Speech model pushed to Hub as Text2Audio

🤗Transformers
1

Im trying to finish the Audio course and after trying to do the third exercise of fine-tuning SpeechT5 model, after pushing the model to Hub, the model is labeled as text-to-audio and thus isnt considered as a correct submission for the course…

I’ve tried:
kwargs = { "dataset_tags": "facebook/voxpopuli", "dataset": "Vox Populi - Lithuanian", "model_name": "speecht5-lt", "finetuned_from": checkpoint, "tasks": "text_to_speech", }

trainer.push_to_hub(**kwargs)

and also just trainer.push_to_hub(), but neither seems to work.

2

I’m having the same issue, did you find a solution?

3

I think it was something really dumb, like using “_” instead of “-” in tasks…

Here is a working kwargs dict that I ended up using successfully:
kwargs = {
“dataset_tags”: “mozilla-foundation/common_voice_13_0”,
“finetuned_from”: checkpoint,
“tasks”: “text-to-speech”,
“tags”: “text-to-speech”,
“language”: “pt”,
}

trainer.push_to_hub(**kwargs)

Hope this helps!